ebook img

Height and Intersection for a family of semi-stable curves PDF

21 Pages·0.28 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Height and Intersection for a family of semi-stable curves

HEIGHT AND ARITHMETIC INTERSECTION FOR A FAMILY OF SEMI-STABLE CURVES 9 SHU KAWAGUCHI 9 9 1 Abstract. In this paper, we consider an arithmetic Hodge index theorem for a family of n semi-stable curves, generalizing Faltings-Hriljac’s arithmetic Hodge index theorem for an a arithmetic surface. J 7 2 ] Introduction G A Inpapers[4]and[7],FaltingsandHriljacindependentlyprovedthearithmeticHodgeindex . theorem on an arithmetic surface. Moriwaki [12] subsequently proved a higher dimensional h t case of Faltings-Hriljac’s arithmetic Hodge index theorem. In this paper, we consider an a m arithmetic Hodge index theorem for a family of semi-stable curves. Namely, we prove [ Theorem A (cf. Theorem 5.2). Let K be a finitely generated field over Q, XK a geo- 1 metrically irreducible regular projective curve over K, and L a line bundle on X with K K v degL = 0. Let B = (B,H) be a polarization of K, i.e., B a normal projective arith- 7 K 2 metic variety with the function field K, and H a nef C∞-hermitian Q-line bundle on B. 1 f 1 Let (X B,L) be a model of (XK,LK) (please see 4 for terminology). We make the −→ § 0 following assumptions on the model: 9 9 (a) f is semi-stable; h/ (b) XC and BC are non-singular and fC : XC BC is smooth. → t a Let J be the Jacobian of X and Θ a divisor on J which is a translation of the theta K K K K m divisor on Picg−1(X ) by a theta characteristic. Then we have K : v i deg c (L)2 c f∗(H) d 2hB ([L ]), X 1 · 1 ≤ − OJK(ΘK) K ar where [LK] denotes thde p(cid:16)obint of JbK(cid:0)corresp(cid:1)on(cid:17)ding tob LK (For the definition of a height function hB , please see 4). OJK(ΘK) § Furthermore, we assume that H is ample and c (H) is positive. Then the equality holds 1 b if and only if L satisfies the following properties: (a) There is a Zariski open set B′′ of B with codim (B B′′) 2 such that deg(L ) = 0 B C \ ≥ | for any fibral curves C lying over B′′. (b) The restriction of the metric of L to each fiber is flat. Date: 19/January/1999,15:30 +0900,(Version 1.0). 1991 Mathematics Subject Classification. 14G40. Key words and phrases. arithmetic variety, Hodge index theorem. Partially supported by JSPS Research Fellowships for Young Scientists. 1 2 SHU KAWAGUCHI We note that when B is the spectrum of the ring of integers, the above theorem is nothing but the arithmetic Hodge index theorem for a semi-stable arithmetic surface. Our proof uses arithmetic Riemann-Roch theorem, similar to that of Faltings on an arith- meticsurface,althoughwemustconsidertheQuillenmetric. Nowweoutlinetheorganization of this paper. In 1, we recall some properties of relative Picard functors. In 2, we recall § § some facts on determinant line bundles, especially for semi-stable curves. In 3, we deal § with an arithmetic setting and give hermitian metrics to the results of 2. In 4, we quickly § § review (a part of) the theory of height functions over a finitely generated field over Q, due to Moriwaki [13]. Finally in 5, we prove the main theorem. § I wish to express my sincere gratitude to Professor Moriwaki for his incessant warm en- couragement. Moreover, it is he who suggested that I consider this work. 1. The Picard functor The purpose of this section is to review some properties of the relative Picard functor, which we will use later. We refer to [2, 8-9] for details. In this section, we only deal with §§ schemes which are locally noetherian. Let S be a locally noetherian base scheme, f : X S a flat, projective morphism. The → relative Picard functor Pic of X over S is the fppf-sheaf associated with the functor X/S P : (locally noetherian S-schemes) (Sets), T Pic(X T). X/S S → 7→ × If we assume f ( ) = holds universally, then for all locally noetherian S-schemes ∗ X S O O g : T S, → Pic (T) = Pic(X T)/Pic(T). X/S S × Furthermore, if X/S admits a section ǫ : S X, then one checks immediately, → group of isomorphism classes of (1.1) Pic (T) = invertible sheaves L on X T, . X/S  ×S  plus isomorphism (ǫ g,1T)∗(L) T ◦ ≃ O Such invertible sheaves are said to be rigidified along the induced section ǫ = ǫ g.   T ◦ IfS consistsofafield, thenPic isagroupscheme. LetPic0 beitsidentitycomponent. X/S X/S For a general locally noetherian scheme S, we introduce Pic0 as the subfunctor of Pic X/S X/S which consists of all elements whose restrictions to all fibers X , s being a point of S, belong s to Pic0 . Xs/k(s) If X is a proper curve over a field k, then Pic0 consists of all elements of Pic whose X/k X/k partial degree on each irreducible components of X k is zero, where k is an algebraic k ⊗ closure of k. We note that if Pic (resp. Pic0 ) is representable by a locally noetherian scheme, X/S X/S then for all locally noetherian S-schemes T, Pic T = Pic (resp. Pic0 T = Pic0 ). X/S ×S X×ST/T X/S ×S X×ST/T Now we introduce the notion of universal line bundles when Pic (resp. Pic0 ) is X/S X/S representable by a locally noetherian scheme. We assume that the structural morphism f : X S admits a section ǫ and that f ( ) = holds universally, so that Pic is ∗ X S X/S → O O HEIGHT AND INTERSECTION 3 given by (1.1) for a locally noetherian S-scheme. If Pic (resp. Pic0 ) is representable X/S X/S by a locally noetherian scheme, then the identity on Pic (resp. Pic0 ) gives rise to a X/S X/S line bundle U (resp. U0) on X Pic (resp. X Pic0 ) which is rigidified along the ×S X/S ×S X/S induced section. U (resp. U0) is called the universal line bundle. The justification of the notion of “universal” is: Proposition 1.1. Let f : X S be a flat morphism of locally noetherian schemes and let → ǫ be a section of f. Assume that f ( ) = holds universally. If Pic (resp. Pic0 ) is ∗ OX OS X/S X/S representable by a locally noetherian scheme, the universal line bundle U has the following property: For every locally noetherian scheme g : T S, and for every line bundle L′ on → X′ = X T which is rigidified along the induced section ǫ′ = ǫ g, there exists a unique S × ◦ morphism g : T Pic such that L′ is isomorphic to (1 g)∗(U). X/S → × If Pic0 is representable by a locally noetherian scheme, the universal line bundle U0 has X/S a similar property for a line bundle L′ on X′ = X T which is rigidified along the induced S × section and L′ Pic0 for all t T. t ∈ Xt/k(t) ∈ Proof. [2, 8.2. Proposition 4] 2 Now we restrict ourselves to the case of semi-stable curves. We recall that a semi-stable curve of genus g is a proper flat morphism f : X S whose fiber X over every geometric s → point s of S is a reduced connected curve with at most ordinary double points such that dim H1(X , ) equals to g. k(s) s OXs Proposition 1.2. Let f : X S be a semi-stable curve of locally noetherian schemes. → Then f ( ) = holds universally. ∗ X S O O Proof. We have only to prove that f ( ) = . Let π f be the Stein factorization of ∗ X S O O ◦ f, where f : X S is a proper morphism with connected fibers and π : S S is a finite → → morphism. Since every fiber is geometrically reduced and geoemetrically connected, there is a section ηe : S Se such that f = η f by rigidity lemma ([14, Proposietion 6.1]). Since → ◦ f ( ) factors through OS ≃ ∗ OX e e e e OS → η∗(OS) → f∗η∗(OS) = f∗(OX), η ( ) is injective. On the other hand, since η is a closed immersion, η ( ) OS → ∗ OS e e OS → ∗ OS is surjective, hence = η ( ). Then, f ( ) = π f ( ) = π ( ) = π (η ( )) = . e OS ∗ OS ∗ OX ∗ ∗ OX ∗ OS ∗ ∗ eOS OS2 e e e Wefinishthissectionbyquoting aresult obtainedbyDeligneconcerning therepresentabil- ity of the relative Picard functor. Theorem 1.3. Let f : X S be a semi-stable curve of locally noetherian schemes. Then → Pic is a smooth algebraic space over S. The identity component Pic0 is a semi-abelian X/S X/S scheme. Proof. [2, 9.4. Theorem 1] or [3, Proposition 4.3] 2 4 SHU KAWAGUCHI 2. Determinant line bundles The purpose of this section is to review some properties of determinant line bundles. Since we are concerned about a family of curves in this paper, we only consider determinant line bundles in a restricted context. For a general treatment of determinant line bundles, we refer to [11]. Theorem 2.1. Let us consider a morphism f : X S of noetherian schemes with the → following conditions: (i) f is proper, f ( ) = , and dimf = 1; ∗ X S O O (ii) There is an effective Cartier divisor D on X such that D is f-ample and flat over S. For every f : X S satisfying the above conditions, for every line bundle L on X and isomorphism of sh→eaves φ : L ∼ L′, one can uniquely construct a line bundle detRf (L) ∗ on S and an isomorphism de−t→Rf (φ) : detRf (L) ∼ detRf (L′) in such a way that ∗ ∗ ∗ −→ detRf (L) becomes a functor with the following properties: ∗ (a) If f (L) and R1f (L) are both locally free, then ∗ ∗ detRf (L) = detf (L) detR1f (L) −1; ∗ ∗ ∗ ⊗ (b) detRf (L) is compatible with a base change(cid:0), i.e., if g :(cid:1) T S is a morphism of ∗ → noetherian schemes, then g∗(detRf (L)) = detR(f ) (L ); ∗ ∼ T ∗ T (c) If S is connected and M is a line bundle on S, then detRf (L f∗(M)) = detRf (L) Mχ, ∗ ∼ ∗ ⊗ ⊗ where χ = χ(C ,L ) for some s S; s s ∈ (d) If D is an effective Cartier divisor on X which is flat over S, then detRf (L) = detRf (L( D)) detf (L ). ∗ ∼ ∗ ∗ D − ⊗ | Proof. [11] or [10, VI 6] 2 § Suppose now that f : X S is a semi-stable curve of noetherian schemes and assume → that f admits a section ǫ. Moreover, let A be a rigidified line bundle on X of degree g 1. − By Theorem 1.3, Pic0 is a semi-abelian scheme and there exists a universal line bundle X/S U0 on X Pic0 . Let Pa be the scheme which is the translation of Pic0 by A, i.e., ×S X/S X/S rigidified line bundle L on X T Pa(T) = . such that L A−1 belongs to Pic0 ( ⊗ X/S) Moreover, let Ua be the line bundle on Pa which is the translation of U0 by A. If qa : X Pa Pa is the second projection, then qa satisfies the condition of Theorem 2.1, S × → because f : X S satisfies the condition of Theorem 2.1. Thus the determinant line bundle → detRqa(Ua) on Pa is defined. To simplify the notation, let us denote detRqa(Ua) by −1. ∗ ∗ T In the following, we will see that −1 is related to the theta divisor. Here we further T assume that f : X S is smooth of genus g 1. First, we define the theta divisor. → ≥ HEIGHT AND INTERSECTION 5 Let (X/S)(g−1) be the symmetric (g 1)-fold product, i.e., − (g−1)times (X/S)(g−1) = X X/S , S S g−1 × ···× z }| ({g−1)times where the (g 1)-th symmetric group S acts on X X naturally. Let g−1 S S − × ···× (X/S)(g−1) Picg−1, Dz }[D| ] { → X/S T → T be a morphism, where for any locally noetherian S-schemes T and for any T-valued point D of (X/S)(g−1) (i.e., for any effective Cartier divisors on X T of degree (g 1)), we T S × − denote by [D ] the element of Picg−1 corresponding to D . The schematic image of this T X/S T morphism, which turns out an effective relative Cartier divisor on Picg−1, is called the theta X/S divisor for X/S and denoted by Θ . X/S Proposition 2.2. Let f : X S be a projective smooth morphism of noetherian schemes → whose geometric fibers are smooth projective curves of genus g 1. We assume the existence of a section. Let Picg−1 be a Picard scheme of degree (g 1) a≥nd U a universal line bundle X/S − on X Picg−1. Then ×S X/S detRq∗(U) ∼= OPicXg−/1S(−ΘX/S), where Θ is the theta divisor for X/S and q : X Picg−1 Picg−1 is the second X/S ×S X/S → X/S projection. Proof. When the base scheme is a point, or an arithmetic surface, this is well-known (cf. [4, 5] or [10, VI Lemma 2.4]). The proof for a general base scheme is similar to that for a § point, as we will see in the following. Let p : X Picg−1 Picg−1 be the first projection. Let D′ be an effective relative ×S X/S → X/S Cartier divisor of sufficiently large degree on X (actually degD′ g is enough) and put ≥ D = p∗(D′). Since H0(X ,U( D) ) = 0 s t − for allpoints t of Picg−1 andthe point s of S lying below t, q (U( D)) = 0 by [6, Corollorary X/S ∗ − II.12.9], and R1q (U( D)) is locally free. Thus, by (a) and (d) of Theorem 2.1, ∗ − detRq (U) = detq (U ) R1q (U( D)) −1. ∗ ∗ D ∗ | ⊗ − Since q (U) is torsion-free and H0(X ,U ) = 0 fo(cid:0)r a general po(cid:1)int t of P, it follows that ∗ s t q (U) = 0. Also, since D Picg−1 is finite, R1q (U ) = 0. Thus we get the exact ∗ → X/S ∗ |D sequence: 0 q (U ) R1q (U( D)) R1q (U) 0. ∗ D ∗ ∗ → | → − → → We denote the homomorphism q (U ) R1q (U( D)) by α. Since R2q (U) = 0, we get ∗ D ∗ ∗ | → − by [6, Theorem II.12.11] R1q (U) k(t) = H1(X ,U ) ∗ ∼ s t ⊗ 6 SHU KAWAGUCHI for all points t of Picg−1 and the point s of S lying below t. If R1q (U) k(t) = 0, then X/S ∗ ⊗ R1q (U) is also zero for some neighborhood of t, and especially R1q (U) is flat for some ∗ ∗ neighborhood of t. Thus α(t) is an isomorphism R1q (U) k(t) = 0 ∗ ⇔ ⊗ H1(X ,U ) = 0 s t ⇔ t Θ . X/S ⇔ 6∈ Therefore if we put E = t Picg−1 (detα)(t) = 0 , then E = aΘ for some positive { ∈ X/S | } X/S integer a. By considering the case that the base scheme is a point, we get a = 1. 2 Now we put everything together and get: Theorem 2.3. Let f : X S be a semi-stable curve of genus g 1 of noetherian schemes → ≥ and assume that f admits a section ǫ. Let A be a rigidified line bundle of degree (g 1) − and (Pa,Ua) the translation of (Pic0 ,U0) by A. We put −1 = detRqa(Ua), where qa : X/S T ∗ X Pa Pa is the second projection. Then, S × → (i) If T S be a morphism of noetherian schemes such that f : X T T is smooth, T S → × → then −1 = ( Θ ) TT OPTa − XT/T where Θ is the theta divisor for X /T. XT/T T (ii) If L is a rigidified line bundle on X which belongs to Pa(S), then there is a canonical morphism ga : S Pa such that the induced morphism → u : detRf (L) (ga)∗( −1) L ∗ → T is canonically isomorphic. Proof. Noting that determinant line bundles are compatible with a base change, we have already seen (i). Regarding as (ii), by the universal property of Ua, there exists a canonical morphism ga : S Pa such that → L = (1 ga)∗(Ua). ∼ × On the other hand, since determinant line bundles are compatible with a base change, we have canonically (ga)∗(detRqa(Ua)) = detRf ((1 ga)∗(Ua)). ∗ ∼ ∗ × 2 Combining above two isomorphisms, we get the desired isomorphism. 3. Arithmetic Setting In this section, we consider an arithmetic setting. An arithmetic variety is an integral scheme which is flat and quasi-projective over Spec(Z). Let f : X B be a semi-stable curve of genus g 1 of arithmetic varieties and assume → ≥ that f admits a section ǫ. We also assume that fC : XC BC is a smooth morphism. Let → A be a rigidified line bundle of degree (g 1) and (Pa,Ua) the translation of (Pic0 ,U0) − X/S by A. We put −1 = detRqa(Ua) on Pa, where qa : X Pa Pa is the second projection. T ∗ × → HEIGHT AND INTERSECTION 7 Then by Theorem 2.3(ii), for a rigidified line bundle L which belongs to Pic0 , we have a X/S natural isomorphism u : detRf (L A) (ga)∗( −1), L ∗ ⊗ → T where ga : S Pa is an induced morphism by L A. → ⊗ In this section we give metrization on the above line bundles, and consider the norm of uL. Let ΘXC/BC be the theta divisor for XC/BC, which is a relative Cartier divisor on PCa = PicgX−C1/BC. Then by Theorem 2.3(i), TC−1 = OPicXg−C1/BC(−ΘXC/BC). In the following, we introduce a metric on OPicXg−C1/BC(−ΘXC/BC). Put J = Pic0XC/BC and let λ : Picg−1 J [D ] (g 1)[ǫ ] XC/BC → T 7→ − T be an isomorphism, where for any BC-scheme T, [ǫT] is the class of the induced section by ǫ. Let Θ0 be the image of Θ by λ. XC/BC XC/BC We need some definitions to proceed. The Siegel upper-half space of degree g, denoted by , is defined by g H = Ω = X +√ 1Y GL (C) tΩ = Ω,Y > 0 . g g H { − ∈ | } Moreover, the symplectic group of degree 2g, denoted by Sp (Z), is defined by g Sp (Z) = S GL (Z) tSJS = J , g 2g { ∈ | } 0 I A B where J = − . An element S = of Sp (Z) acts on by I 0 C D g Hg (cid:18)− (cid:19) (cid:18) (cid:19) S Ω = (AΩ+B)(CΩ+D)−1 · and Sp (Z) becomes a coarse moduli of principally polarized abelian varieties. g g \H For z = x+√ 1y Cg and Ω = X +√ 1Y , we define g − ∈ − ∈ H θ(z,Ω) = exp(π√ 1 tmΩm+2π√ 1 tm z), − − · m∈Zg X θ (z,Ω) = √4 detY exp( πtyYy) θ(z,Ω) . k k − | | Then θ becomes a holomorphic function on Cg . Moreover θ becomes a C∞-function g ×H k k which is periodic with respect to Zg+ΩZg, so that θ is seen as a C∞-function on Cg/Zg+ k k ΩZg. Going back to our situations, for any b B(C), let us write analytically ∈ J = Cg/Zg +Ω Zg b ∼ b where Ω . Then there is a unique element t Cg/Zg + ΩZg such that Θ0 = b ∈ Hg b ∈ Xb div(θ(z +t ,Ω )), where θ(z +t ,Ω ) is seen as a function of z. b b b b Proposition 3.1. Let the notation be as above. Let 1 denote the section of (Θ0 ) OJ XC/BC which corresponds to Θ0 . For any p J, let b B(C) be the point lying below p and XC/BC ∈ ∈ write J = Cg/Zg +Ω Zg and Θ0 = div(θ(z +t ,Ω )) with t Cg/Zg +ΩZg. Moreover, b ∼ b Xb b b b ∈ let z Cg/Zg +ΩZg correspond to p. Then, if we define ∈ 1 Θ0 (p) = θ (z +tb,Ωb), k k XC/BC k k 8 SHU KAWAGUCHI then k·kΘ0XC/BC gives a C∞ metric on OJ(Θ0XC/BC) Proof. If the base space B(C) is a point, the assertion is well-known (cf. [4, 3]). Thus § all we need to prove is that k1kΘ0XC/BC varies smoothly as b ∈ B(C) varies. However, since the morphism Φ : B(C) Sp (Z) , b the class of J g g b → \H 7→ is holomorphic and tb is given the difference of the section ǫC and a theta characteristic, 1 Θ0 varies smoothly as b B(C) varies. 2 k k XC/BC ∈ Finally, OPicgX−C1/BC(ΘXC/BC) is metrized by OJ(Θ0XC/BC),k·kΘ0XC/BC through λ. We write this metric by . (cid:16) (cid:17) Θ k·k XC/BC Next we give a C∞ metric on LC over XC. Actually, there is a certain class of C∞ metrics on LC which is suitable for our purpose. We introduce this class in the following. First we recall admissible metrics of line bundles on a compact Riemann surface. Let M be a compact Riemann surface of genus g 1 and ω ,ω , ,ω a basis of H0(M,Ω1 ) ≥ { 1 2 ··· g} M with √ 1 − ω ω = δ . i j ij 2 ∧ ZM Let us put g √ 1 µ = − ω ω . i i 2g ∧ i=1 X Then µ is a positive (1,1)-form on M, and is called the canonical volume form on M. A C∞-metric h of a line bundle L on M is said to be admissible if L c ((L,h )) = (degL)µ. 1 L For every line bundle on M, we can endow an admissible metric unique up to a constant multiplication. Now let us go back to our situation, i.e., the case that f : XC BC is a smooth family → of curves of genus g 1. A C∞-metric hL on LC over XC is said to be admissible if for any ≥ b B(C), its restriction (L ,h ) on X is admissible. The existence of an admissible metric b L,b b ∈ is given by: Proposition 3.2. Let X and B be smooth varieties over C and f : X B a smooth → projective morphism with a section whose fibers are curves of genus g 1. Let L be a line ≥ bundle on X. Then there exists a (global) admissible metric on L over X. Proof. First we construct a suitable (1,1)-form on X. Let j : X J = Pic0 → X/B is the embedding induced by the section. On J, we have a C∞-hermitian line bundle OJ(Θ0X/B),k·kΘ0X/B by Proposition 3.1. We consider (cid:16) (cid:17) 1 ω = gj∗ c1 OJ(Θ0X/B),k·kΘ0X/B . (cid:16) (cid:16) (cid:17)(cid:17) HEIGHT AND INTERSECTION 9 Then, for any b B, ω = ω is the canonical volume form on X (cf. [4, Thoerem 1]). ∈ b |Xb b Next we note that the statement of the proposition is local with respect to B. Indeed, let U ∞ be an open covering of B and assume that for each i there is an admissible metric { i}i=0 h(Li) on L|f−1(Ui) over f−1(Ui). Take a partition of unity {ρi} subordinate to {Ui}. Then it is easy to see that ∞ f∗(ρ )h(i) i L i=0 X is an admissible metric on X. For b B, we consider a small ball U B containing b. We set X = f−1(U). Then U ∈ ⊂ ≈ there is a diffeomorphism g : X X U over U ([9, Theorem 2.4]). U b −→ × Let h be any C∞-hermitian metric on L over X and set η = c (L,h ). Then η is a d- L U 1 L closed (1,1)-form. We claim that (deg(L)ω η) is d-exact. Indeed, (deg(L)ω η) = − |XU − |Xb 0 in H2(X ,C). On the other hand, H2(X ,C) = H2(X ,C) by Poincar´e’s lemma. Thus b U b (deg(L)ω η) is a d-exact (1,1)-form. Then by the ddc-lemma, there is a C∞ function − |XU ψ on X with ddc(ψ) = (deg(L)ω η) . Now if we set h′ = exp( ψ)h on L U − |XU L,U − L|XU |XU over X , then U ddc c (L ,h′ ) = deg(L)ω , 1 |XU L,U |XU which is an admissible metric on(cid:0)L|XU over XU(cid:1). 2 Now we prove the main proposition of this section, which will be a key point to prove Proposition 5.1. Proposition 3.3. Let f : X B be a semi-stable curve of genus g 1 of arithmetic → ≥ varieties and assume that f admits a section ǫ. We also assume that fC : XC BC is a → smooth morphism. Let u : detRf (L A) (ga)∗( −1), L ∗ ⊗ → T be the isomorphism given at the beginning of this section. We endow C∞ metrics on A and ω , an admissible metric on L, and then the Quillen metric on detRf (L A) determined X/B ∗ ⊗ by these metrics. Moreover, we endow a metric −1 on −1. Then the norm of u is k·kΘXC/BC T L independent of L. Proof. Let b B(C). Since determinant line bundles are compatible with a base change ∈ and since the Quillen metric is given fiberwise, we get uL : detRfb∗(Lb ⊗Ab) → OPicXg−b1(−ΘXb)|[Lb⊗Ab], where [L A ] is the point corresponding to L A on Picg−1. Then by the following b ⊗ b b ⊗ b Xb 2 lemma, we obtain Proposition 3.3. Lemma 3.4. Let M be a compact Riemann surface of genus g 1, L a line bundle of degree ≥ 0 on M. We endow a C∞-metric hA on A, a C∞-metric hΩ1 on Ω1M, and an admissible M metric h on L. Then we have a canonical isomorphism L uL : detΓ(L⊗A) → OPicMg−1(−ΘM)|[L⊗A], 10 SHU KAWAGUCHI where detΓ(L A) is the determinant line bundle of L A. We endow the Quillen metric ⊗ ⊗ on detΓ(L⊗A) and k·k−Θ1M on OPicMg−1(−ΘM). Then the norm of uL is independent of L. Proof. Let h′ and h′ be admissible metrics on A and Ω1 respectively. We write the A Ω1 M M Quillen metric defined by (L⊗A,hL⊗hA) and (Ω1M,hΩ1M) as hQL⊗A. We also write the Quillen ′ metrics defined by (L A,h h′ ) and (Ω1 ,h′ ) as hL⊗A . We decompose u into ⊗ L ⊗ A M Ω1M Q L detΓ(L A),hL⊗A α detΓ(L A),hL⊗A′ ⊗ Q −→ ⊗ Q (cid:16) (cid:17) (cid:16) −β→ detΓ(cid:17)(L⊗A),hFL⊗A −γ→ OPicMg−1(ΘM)|[L⊗A], where hL⊗A is the Faltings’ metric on L A(cid:0). By the definition (cid:1)of the Quillen metrics, the F ⊗ norm of α is independent of L, because we only change the metric of A. The norm of β is the difference of the Quillen metric and the Faltings’ metric for admissible line bundles, which is a constant depending only on M (cf. [15, 4.5]). Moreover, the norm of γ is also independent of L, which is actually given by exp(δ(M)/8) with the Faltings’ delta function δ(M) (Or rather, this is the definition of δ(M)) . Therefore the norm of u is independent L of L. 2 4. Arithmetic height functions over function fields A. Moriwaki [13] has recently constructed a theory of arithmetic height functions over function fields, with which he recovered the original Raynaud theorem (i.e., over a finitely generated field over Q). In this section, we see a part of his theory. Let K be a finitely generated field over Q with tr.deg (Q) = d. Let B be a normal K projective arithmetic variety with the function field K. Let H be a nef C∞-hermitian Q-line bundle on B, i.e., deg(H ) 0 for any curve C and c (H) is semi-positive on B(C). A pair C 1 | ≥ B = (B,H) with the above properties is called a polarization of K. Moreover, we say that a polarization B isdbig if rkH0(B,H⊗m) grows the order of md and that there is a non-zero section s of H0(B,H⊗n) with s < 1 for some positive integer n. sup k k Let X be a projective variety over K and L a line bundle on X . By a model of K K K f (X ,L ) over B, we mean a pair (X B,L) where f : X B is a projective morphism K K −→ → of arithmetic varieties and L = (L,h ) is a C∞-hermitian Q-line bundle on X such that, on L the generic fiber, X and L coincide with X and L respectively. K K f By abbreviation, a model (X B,L) is sometimes written as (X,L). We note that −→ although we use the notation X and L , a model of (X ,L ) is not a priori determined. K K K K For P X(K), we denote by ∆ the Zariski closure of the Image Spec(K) X in X. P K ∈ → f Then we define the height of P with respect to (X B,L) to be (cid:0) (cid:1) −→ 1 hB (P) = deg c (L ) c (f∗H )d . (X,L) [K(P) : K] 1 |∆P · 1 |∆P (cid:0) (cid:1) If we change models of (X ,L ), then heigdht functions differ by only bounded functions on K K b b X (K). Namely, if (X,L) and (X′,L′) are two models of (X ,L ), then there is a constant K K K

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.